In this paper. a new pole zero order determination procedure is proposed for ECG signals. The method is based upon the singular value decomposition of the reduced rank approximation of the data matrix obtained by linear filtering of the DCTA coefficients of the ECG signal. The energy characteristics of the reduced rank components are found to be highly order dependent. In this paper, we have made use of this property to identify the order of the underlying model. This NMSE of the waveforms reconstructed with the orders estimated using the proposed method are found to be lower than those obtained using statistical methods such as the AIC.
展开▼